search

msakarvadia / llm_bias

Investigating if we can find circuits in LLMs that reinforce human-biases found in training data
MIT License
0 stars 0 forks source link

readme

llm_bias

Investigating if we can find circuits in LLMs that reinforce human-biases found in training data

Installation

Requirements: python >=3.7,<3.11

git clone https://github.com/msakarvadia/llm_bias.git
cd llm_bias
conda create -p env python==3.10 (or conda create --prefix=env python=3.10)
conda activate env
pip install -r requirements.txt

Black

To maintain consistent formatting, we take advantage of black via pre-commit hooks. There will need to be some user-side configuration. Namely, the following steps:

  1. Install black via pip install black (included in requirements.txt).
  2. Install pre-commit via pip install pre-commit (included in requirements.txt).
  3. Run pre-commit install to setup the pre-commit hooks.

Once these steps are done, you just need to add files to be committed and pushed and the hook will reformat any Python file that does not meet Black's expectations and remove them from the commit. Just re-commit the changes and it'll be added to the commit before pushing.

Getting interactive node on perlmutter

salloc --nodes 1 --qos interactive --time 01:00:00 --constraint gpu --gpus 4 --account=mxxxx